AdaBoost Face Detection Based On Locally Adaptive Kernel Regression

As the key to face recognition and face image information processing, the result of face detectionwill directly affect the follow research projects. There are many difficult in the process of applyingthis technology, mainly because images are vulnerable to the impact of imaging equipment, imagingconditions and the way of storage. The research in this area now has become a complex andchallenging research topic.This article describes the basic concepts of face detection technology, analysis and presents thebasic theory of several face detection methods, basing on that to explore the face detection process offeature extraction, classification and composition of the detection system. This paper discussesperformance about AdaBoost face detection and locally adaptive kernel regression (LARK) featureextraction method and combines these algorithms to constitute a face detection system in order toimprove the accuracy of face detection. This research work includes the following aspects:First, studying the face detection algorithm based on AdaBoost. We discuss the three parts of facedetection system which is made up of Haar features, integral image and cascade classifier based onAdaBoost algorithm. Aiming at the weak ability problem on each classifier may undetected, weexperiment and analysis this phenomenon.Second, studying the feature extraction method based on locally adaptive kernel regression. Thismethod is very effective to capture the local structure of the data. It is used to describe the edgeposition and the shape of graphics in the image. It constitutes a face detection system. Theexperiments about this system show the advantages of this method to depict the image information indetection process. But the test results are influenced by similar regions in the target image.Finally, we working out the face detection algorithm based on the feature extraction methodabout LARK and AdaBoost algorithm. To improve the detection rate of system, using the test resultsof each classifier as input to the classifier based on the LARK feature extraction method. We compareand verify the effect of the face detection algorithm through the experiment.The article summarizes the paper at last and predicts the development trend about this topic.